English

Detecting Deep Neural Network Defects with Data Flow Analysis

Machine Learning 2019-10-01 v2 Signal Processing Machine Learning

Abstract

Deep neural networks (DNNs) are shown to be promising solutions in many challenging artificial intelligence tasks. However, it is very hard to figure out whether the low precision of a DNN model is an inevitable result, or caused by defects. This paper aims at addressing this challenging problem. We find that the internal data flow footprints of a DNN model can provide insights to locate the root cause effectively. We develop DeepMorph (DNN Tomography) to analyze the root cause, which can guide a DNN developer to improve the model.

Keywords

Cite

@article{arxiv.1909.02190,
  title  = {Detecting Deep Neural Network Defects with Data Flow Analysis},
  author = {Jiazhen Gu and Huanlin Xu and Yangfan Zhou and Xin Wang and Hui Xu and Michael Lyu},
  journal= {arXiv preprint arXiv:1909.02190},
  year   = {2019}
}

Comments

2 pages

R2 v1 2026-06-23T11:06:13.493Z